[gnugo-devel] Code removal (#25)

[Gunnar Farnebäck]

In http://lists.gnu.org/archive/html/gnugo-devel/2003-07/msg00059.html
I proposed removal of certain obsolete pieces of code to improve
maintenance after the release of 3.4. Now I think it's time for
another round. This time I propose removal of the items below. Please
protest if you disagree or add things I have missed.

* experimental pattern based tactical reading

This was left in the post 3.4 cleaning but have since then received no
attention. It's also way slower than the usual tactical reading
(several times slower, possibly as much as eight) and has a big
negative breakage.


* tree-based pattern matching

Was also left in the post 3.4 cleaning but has received no attention
since. It's also rather intrusive, in particular in mkpat.c.


* debugboard

Was left in the post 3.4 cleaning awaiting a replacement. Now
view.pike is quite useful, while debugboard in the meantime has broken
from a lack of maintenance.


* alternate connections

Is there any point in keeping those around any longer? Was also left
in the post 3.4 cleaning but nobody seems much interested.


* cutstone.tst, semeai2.tst, and *_rot.tst

I proposed removing these in
http://lists.gnu.org/archive/html/gnugo-devel/2004-11/msg00069.html
which went without response and I forgot about it at the time.


* unused game records in regression/golois

The regression/golois directory contains sgf files provided by Tristan
Cazenave together with the test cases in capture.tst, connect.tst,
vie.tst, and global.tst. Only a third of the game records in
regression/golois are used in any tests, however. The rest occupies
space without being of any use and are not particularly interesting
at all from a GNU Go point of view.


* useless information in kgs game records

As you may have noticed game records from kgs are rather verbose, e.g.
with lines like:
;B[pq]CR[pq]BL[288.459]OB[22]
where the first property (the move) is meaningful, the second property
(a circle at the move) is utterly useless and the last two properties
(timing information) are at best marginally useful in the regression
context. 


* transpat and patlib

Old, unused, and broken code to transform patterns in *.db files.

/Gunnar